Welcome, guest | Sign In | My Account | Store | Cart

Generate classes with named data attributes that can be sequenced. Useful for POD classes for which many records will exist concurrently.

Compare the feature set to NamedTuples by Raymond Hettinger: http://code.activestate.com/recipes/500261-named-tuples/

Python, 326 lines
  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
#! /usr/bin/env python
######################################################################
#  Written by Kevin L. Sitze around 2008-05-03
#  This code may be used pursuant to the MIT License.
######################################################################

"""Generate classes with named data attributes that can be sequenced.
Useful for POD classes for which many instances will exist.
Compare this feature set to NamedTuples by Raymond Hettinger:
http://code.activestate.com/recipes/500261-named-tuples/

>>> Point = NamedSequences('Point', 'x', 'y')
>>> Point                               # module.class
<class '__main__.Point'>
>>> Point.__doc__                       # documentation
'Point(x, y) => instance'
>>> Point()                             # default fields are undefined
Point(x = None, y = None)
>>> Point = NamedSequences('Point', 'x', y=0) # specify new defaults
>>> Point()                             # default fields take our values
Point(x = None, y = 0)
>>> Point(1, 2)                         # positional parameters
Point(x = 1, y = 2)
>>> p = Point(y=3, x=4)                 # keyword parameters
>>> p
Point(x = 4, y = 3)
>>> p.x                                 # member access and...
4
>>> p[0]                                # ...indexing and...
4
>>> p[1]
3
>>> p[2]                                # ...bounds checking
Traceback (most recent call last):
  [snip]
    IndexError: tuple index out of range
>>> x, y = p                            # unpack on assignment
>>> x, y
(4, 3)
>>> d = p._asdict()                     # as dictionary
>>> d
{'y': 3, 'x': 4}
>>> Point(**d)                          # from dictionary
Point(x = 4, y = 3)
>>> p._replace(x=10)                    # replace fields by name
Point(x = 10, y = 3)
"""

import keyword
import copy
import sys

__all__ = ( 'NamedSequences', 'named_sequences' )

def unique( it ):
    """unique( it ) => iterator

    Generate each value from the input iterator "it" exactly once
    ordered according to the first occurance of the value.
    """
    seen = set()
    for v in it:
        if v in seen: continue
        seen.add( v )
        yield v

def NamedSequences( className, *_names, **_kwds ):
    """NamedSequences( className[, 'name_1'[,...[,
                       name_N = defaultValue_1[, ... ]]]] )

    Construct a new class that contains only the named elements.
    This is useful in cases where you know you're going to have a ton
    of instances for a class and wish to conserve memory by avoiding
    the overhead of a per instance dictionary.

    Instances of the returned class are Python sequences.

        NamedSequences(
            __name__, 'MyClass',
            'fieldName1', ..., # argument field names
            fieldNameN = defaultValue1, ...)

    Keyword arguments are unordered dict's so if you wish to control
    field ordering you MUST specify your field names twice, once in
    the positional argument list area (to define the ordering), and
    the second time in the keyword area to specify that field's
    default value.  Field names defined only as keyword arguments are
    placed in ascending order after all field names specified as
    positional arguments.  Duplicate field names are dropped with
    priority given to the first appearance of the name.

    Keyword arguments beginning and ending with a double underscore
    are Python reserved words.  These will be inserted directly into
    the class __dict__ rather than added as regular field names.

    For example you can redefine the module that the class belongs to
    using the following pattern:

        Point = NamedSequences('Point', 'x', 'y', __module__ = __name__ )

    though this is actually redundant as the default is the module
    from which this function was called.  You can even specify your
    own documentation for the class using "__doc__ = '<my_docs>'".
    """
    # versatile arguments: (klass, 'x y'), (klass, 'x', 'y') or (klass, 'x,y')
    if len( _names ) == 1 and isinstance( _names[0], basestring ):
        _names = _names[0].replace( ',', ' ' ).split()
    _names = tuple( map( str, _names ) )

    def is_identifier( s ):
        try:
            class Tmp( object ): __slots__ = (s,)
        except TypeError: return False
        else: return not keyword.iskeyword( s )

    if not is_identifier( className ):
        raise ValueError( 'class name "%s" is not a valid identifier' % className )
    for name in _names + tuple( _kwds.keys() ):
        if not is_identifier( name ):
            raise ValueError( 'field name "%s" is not a valid identifier' % className )

    # Extract Python reserved words
    extras = dict( ( k, v ) for k, v in _kwds.iteritems() if k.startswith( '__' ) and k.endswith( '__' ) )
    _kwds = copy.deepcopy( _kwds )                 # no messing with defaults
    for k, v in extras.iteritems(): _kwds.pop( k ) # remove reserved words
    try: extras.setdefault( '__module__', sys._getframe( 1 ).f_globals.get( '__name__', '__main__' ) )
    except ( AttributeError, ValueError ): pass # no getting at the module in this env...
    _slots = tuple( unique( _names + tuple( sorted( _kwds.keys() ) ) ) ) # there can be only one...

    class SequenceClass( object ):
        """This class is the superclass of the class generated by
        "NamedSequences".  Each call to "NamedSequences" creates a
        new "SequenceClass" superclass.
        """
        __slots__ = ()
        def __init__( self, *argv, **kwds ):
            for n, v in zip( _slots, argv ): setattr( self, n, v )   # apply positional args 1st
            for n, v in kwds.iteritems(): setattr( self, n, v )     # apply keyword args 2nd
        def __cmp__( self, other ):
            if self.__class__ is other.__class__:
                return cmp( tuple( self ), tuple( other ) )
            else:
                raise TypeError, "requires a '%s' object but received a '%s'" % ( self.__class__.__name__, other.__class__.__name__ )
        def __getattr__( self, n ):
            """Defer applying class defaults until they are actually needed"""
            try: result = _kwds[n]
            except KeyError:
                if n in _slots:
                    result = None
                else: raise
            object.__setattr__( self, n, result )
            return result
        def __getitem__( self, i ): return getattr( self, _slots[i] )
        def __getnewargs__( self ): return tuple( self )
        def __setitem__( self, i, v ): setattr( self, _slots[i], v )
        def __iter__( self ): return iter( getattr( self, n ) for n in _slots )
        def __len__( self ):  return len( _slots )
        def __repr__( self ): return self.__class__.__name__ + '(' + ', '.join( n + ' = ' + repr( getattr( self, n ) ) for n in _slots ) + ')'
        def __str__( self ):  return self.__class__.__name__ + '(' + ', '.join( repr( getattr( self, n ) ) for n in _slots ) + ')'
        def _asdict( self ):  return dict( ( n, getattr( self, n ) ) for n in _slots )
        def _replace( self, **kwds ): return type( self )( *self, **kwds )

    klass = SequenceClass
    extras.setdefault( '__doc__', '%s%s => instance' % ( className, repr( _slots ).replace( "'", "" ) ) )
    extras.update( {
        '__name__': className,
        '__slots__': _slots
    } )
    return type( klass )( className, ( klass, ), extras )

def named_sequences(func):
    """Decorate a function definition to create the tuple.

    @named_sequences
    def Point(x, y): pass
    """
    return NamedSequences( func.__name__, *func.__code__.co_varnames )

if __name__ == '__main__':

    import traceback
    def assertEquals( exp, got ):
        """assertEquals(exp, got)

        Two objects test as "equal" if:
        
        * they are the same object as tested by the 'is' operator.
        * either object is a float or complex number and the absolute
          value of the difference between the two is less than 1e-8.
        * applying the equals operator ('==') returns True.
        """
        from types import FloatType, ComplexType
        if exp is got:
            r = True
        elif ( type( exp ) in ( FloatType, ComplexType ) or
               type( got ) in ( FloatType, ComplexType ) ):
            r = abs( exp - got ) < 1e-8
        else:
            r = ( exp == got )
        if not r:
            print >>sys.stderr, "Error: expected <%s> but got <%s>" % ( repr( exp ), repr( got ) )
            traceback.print_stack()

    def assertException( exceptionType, f ):
        """Assert that an exception of type \var{exceptionType}
        is thrown when the function \var{f} is evaluated.
        """
        try: f()
        except exceptionType: assert True
        else:
            print >>sys.stderr, "Error: expected <%s> to be thrown by function" % exceptionType.__name__
            traceback.print_stack()

    def assertFalse( b ):
        """assertFalse(b)
        """
        if b:
            print >>sys.stderr, "Error: expected value to be False"
            traceback.print_stack()

    def assertTrue( b ):
        if not b:
            print >>sys.stderr, "Error: expected value to be True"
            traceback.print_stack()

    ####
    #  Test NamedSequences
    ####

    Point = NamedSequences( 'Point', 'x', 'y' )
    assertEquals( "<class '__main__.Point'>", repr( Point ) )
    assertEquals( 'Point(x, y) => instance', Point.__doc__ )
    p = Point( 0, 1 )
    q = Point( y = 0, x = 1 )
    assertEquals( 0, p[0] )
    assertEquals( 0, p.x )
    assertEquals( 1, p[1] )
    assertEquals( 1, p.y )
    assertEquals( 1, q[0] )
    assertEquals( 1, q.x )
    assertEquals( 0, q[1] )
    assertEquals( 0, q.y )
    assertException( IndexError, lambda: p[2] )
    x, y = p
    assertEquals( p[0], x )
    assertEquals( p[1], y )
    x, y = q
    assertEquals( q[0], x )
    assertEquals( q[1], y )
    d = p._asdict()
    r = Point( **d )
    assertEquals( p, r )
    s = r._replace( x = 5 )
    assertEquals( 5, s.x )
    assertEquals( r.y, s.y )
    s = s._replace( y = 6 )
    assertEquals( 5, s.x )
    assertEquals( 6, s.y )

    assertEquals( 'Point', Point.__name__ )
    assertEquals( 'Point', p.__class__.__name__ )

    p = Point()
    q = Point()

    assertEquals( q, p )
    assertEquals( 0, cmp( p, q ) )

    p.x = 1
    q.x = 2
    assertTrue( p < q )
    assertTrue( q > p )
    assertTrue( cmp( p, q ) < 0 )
    assertTrue( cmp( q, p ) > 0 )
    p.x = 1
    q.x = 1
    assertEquals( 0, cmp( p, q ) )
    assertEquals( 0, cmp( q, p ) )
    p.y = 1
    q.y = 2
    assertTrue( cmp( p, q ) < 0 )
    assertTrue( cmp( q, p ) > 0 )

    assertEquals( 'Point(1, 1)', str( p ) )
    assertEquals( 'Point(x = 1, y = 1)', repr( p ) )
    assertEquals( p, Point( x = 1, y = 1 ) )

    p.x = '1'
    assertEquals( "Point('1', 1)", str( p ) )
    assertEquals( "Point(x = '1', y = 1)", repr( p ) )

    Point = NamedSequences( 'Point', x = 1, y = 2 )
    p = Point()
    assertEquals( 1, p.x )
    assertEquals( 2, p.y )

    Point = NamedSequences( 'Point', 'x y' )
    p = Point()
    p.x = 0
    p.y = 1

    Point = NamedSequences( 'Point', 'x,y' )
    p = Point()
    p.x = 0
    p.y = 1

    assertException( KeyError, lambda: p.z )
    assertException( AttributeError, lambda: setattr( p, 'z', 1 ) )

    Point = NamedSequences( 'Point', x = 0, y = 0, __module__ = '__name__' )
    assertEquals( "<class '__name__.Point'>", repr( Point ) )

    assertException( ValueError, lambda: NamedSequences( 'for' ) )
    assertException( ValueError, lambda: NamedSequences( 'in' ) )
    assertException( ValueError, lambda: NamedSequences( '123' ) )
    assertException( ValueError, lambda: NamedSequences( 'test', 'for' ) )
    assertException( ValueError, lambda: NamedSequences( 'test', 'in' ) )
    assertException( ValueError, lambda: NamedSequences( 'test', '123' ) )

    @named_sequences
    def Point(x, y, z): pass

    p = Point()
    p.x = 0
    p.y = 1
    p.z = 2

So I stumbled across Raymond Hettinger's Named Tuples recipe recently:

http://code.activestate.com/recipes/500261-named-tuples/

A while ago I had need of something with the following properties:

  1. access to attribute data by name
  2. access to attribute data by index
  3. data can be iterated through (i.e., sequence)
  4. reduced memory use for large numbers of data records
  5. member access (i.e., x.name) to data

The behavior of classes using '__slots__' was not quite good enough for my needs; I needed the sequencing and indexing capabilities too. The result was something that acted like both an object with member variables and a fixed length list.

So, I dug this stuff out of the heap of other descriptor utilities, added some polish to the comments, shamelessly grabbed some of the cool ideas from his recipe (_asdict, _replace, identifier testing, the feature demo and the decorator) and here you go.

Why should you care?

Well, Raymond's recipe is directed at naming fields in a tuple, hence you're dealing with a read only object. This one gives you something that acts like a regular object. You can both look at the fields and change them.

This recipe also provides a couple of things I have found useful:

  • you can muck with the documentation on the object being generated.
  • you may specify default values for any of the fields in the class.
  • you may control which module the class belongs to (useful for higher level factories).
  • the unit tests so you can see some of the things you can do with it.

As a side note, the 'unique' function was garnered off some unknown corner of the net, I don't remember who wrote it originally, and the is_identifier function is based on Zoran Isailovski's recipe here:

http://code.activestate.com/recipes/413487-quick-test-if-strings-are-identifiers/